Coach Amp GmbH
Akazienstrasse 3A
10823 Berlin
info@coachamp.com
Akazienstrasse 3A
10823 Berlin
info@coachamp.com
Coach Amp uses artificial intelligence in a small number of clearly defined places. This page explains where, what it does with your information, what it never does, and the choices you have. It is written to be read - not to be technically complete - and we keep it current as the product changes.
One distinction runs through everything below. If you work with a coach on Coach Amp, anything AI helps produce for you passes through your coach first - they generate it, review it, edit it, and decide whether you ever see it. Separately, if you use Coach Amp on your own, you can choose to generate an insight on your own reflection; that one is yours, by your request, and is not coach-reviewed. We are explicit about which is which.
Helping your coach prepare. Before a session, your coach can generate a briefing that draws together their own notes, your shared reflections and any material you have both agreed to include. Your coach reads and edits this before it informs your session.
Preparing what your coach shares with you. After a session, your coach can generate a summary of what you explored and what you agreed. Your coach reviews and edits it, and nothing reaches you until they publish it.
An assistant for your coach. Coaches can ask questions about their own practice and their work with you, and get suggestions back. This is a thinking aid for your coach. It never contacts you and never acts on its own.
Insights on a reflection you complete on your own. If you complete a reflection yourself - not as part of coaching with a coach - you can ask for a short written interpretation of your results. You choose to generate it, each time, after agreeing once. It appears in the page with a visible "AI-generated" label; it is written by AI and is not reviewed by a coach, and we tell you so at the moment you generate it. This option is never offered inside a coaching relationship - there, your coach is always the one who prepares and reviews what you see.
Comparing a reflection with another person. If you and another person both complete the same reflection, you can generate a written comparison. Both of you agree to this before it happens, and it describes patterns rather than telling either of you what to do.
Drafting text you send. When you share a reflection, you can ask for a suggested message. It is a draft: you read it, edit it and decide whether to send it.
Meeting transcription. If you and your coach use session recording, a third-party service produces the transcript. We retrieve it and it is deleted at the source.
360° feedback carries no AI. When several people give you feedback, the result is arithmetic - averages across the people who responded - and nothing more. No AI writes any part of it.
A small number of reflections have no AI at all. Some reflections deal with sensitive personal transitions. On these, the option to generate AI insights is not offered, and the feature is switched off for them on our servers as well as hidden in the app. We would rather offer nothing than offer an automated interpretation of something that deserves a human conversation.
Benchmarks are statistics, not AI. Where we show you how your results compare with others, that is an average calculated from a group of at least 50 people. No AI is involved and no individual's results are ever shown.
Coach matching is not AI. When we suggest coaches, we apply fixed rules we wrote ourselves. There is no model making that judgement.
AI in Coach Amp does not make decisions about you. It does not decide whether you can use the service, what you pay, which coach you get, or anything else with a consequence attached. It produces text for a person to read and judge.
In a coaching relationship, a person is always in the loop. Everything your coach shares with you passes through your coach first. They generate it deliberately, they can edit it, and they confirm they have reviewed it before publishing. Because your coach reviews this content and takes responsibility for it, we treat them as its author; we do not separately tag it as "AI-assisted", because the human review - not a label - is the safeguard. We tell you plainly, when you start and in your settings, that AI-assisted content your coach shares has been reviewed by them before you see it.
The one exception is yours by request. The insight you can generate on a reflection you complete on your own - outside any coaching relationship - is not reviewed by a coach before you read it. That is the point of it: it is yours, generated when you ask for it. It is shown with a visible "AI-generated" label, and when you generate it we tell you it has not been coach-reviewed.
For an insight on a reflection you complete yourself, we send the numeric ratings you gave and the names of the dimensions and the reflection. That is all. Your name, your email address, your written notes and anything from your other reflections are not included.
For your coach's preparation and summaries, more context is involved - their notes, your shared reflections, and documents you have both agreed to - because that is what makes the output useful to them. Before anything is sent, we remove direct identifiers such as names, email addresses and phone numbers, and we replace names with neutral placeholders.
We never send special categories of personal data to an AI model for profiling, and our systems are instructed never to infer health, sexuality, religion, political opinions or disability from anything you write.
AI processing for Coach Amp takes place within the European Union, in AWS data centres in the Frankfurt region, using Anthropic's Claude models made available through Amazon Bedrock.
That means two companies are involved as processors: Amazon Web Services, which provides the infrastructure and the region in which the model runs, and Anthropic, which builds the model. Both are engaged under data-processing agreements. If you use session recording, the transcription provider is a further processor, listed with the others in our privacy policy.
Nothing you or your coach puts into Coach Amp is used to train, fine-tune or improve any AI model - not ours, and not our providers'. This is a contractual commitment from our model providers, not a setting.
How long AI content is kept: insights and comparisons are kept for as long as you keep the reflection they belong to, and are removed when that reflection - or your account - is deleted (see §6). Working drafts and coach-assistant conversations are removed after 90 days. Technical records that we generate to run the service - which model answered, how long it took, what it cost - are anonymised after 12 months.
AI-generated text can be wrong, bland, or simply miss the point. It reflects the numbers and words it was given and nothing else - it does not know you. Treat it as a prompt for your own thinking, and tell us when it misses: there is a feedback control on every AI insight.
Coach Amp is a coaching and self-development service. Nothing it produces is medical, psychological or therapeutic advice, a diagnosis, or an assessment of your health, and it is not a substitute for professional care. We hold every AI output to the same deliberate standard of language, whether your coach prepared it or you generated it yourself: it does not use clinical or diagnostic vocabulary, it does not present itself as detecting anything about you, and it does not frame you in the language of workplace performance evaluation. This is enforced automatically on every AI output we produce, not left to good intentions.
Write to info@coachamp.com. Use the same address to ask us to delete a reflection, delete your account, or send you a copy of your data. If you want the detail behind this page - how we classify each AI function and why - ask, and we will share what is appropriate.
From 2 August 2026, Article 50 of the EU AI Act requires transparency from anyone providing or professionally using certain AI systems: people must be told when they are interacting with AI, and AI-generated content must be disclosed as such. The safeguards described on this page - visible labelling, a named human reviewing and taking responsibility before anything reaches a client, and no AI inside the coaching conversation itself - are the same safeguards those duties describe. We built them before the deadline, not because of it.
The International Coaching Federation's 2025 Code of Ethics asks something similar of coaches themselves: to disclose their use of AI - including transcription, storage and analysis tools - and to protect client confidentiality through any technology they use. Coach Amp gives coaches the records to answer that cleanly: documented, withdrawable consent, an audit trail of who agreed to what and when, and outputs that are labelled and reviewed. What a coach tells their clients remains their professional responsibility; our job is to make the honest answer easy.
Last updated: 30 July 2026
